New Zealanders alcohol consumption patterns across the lifespan

Transcription

1 New Zealanders alcohol consumption patterns across the lifespan September 2018

2 Prepared for the Health Promotion Agency by: Massey University-the University of Auckland Research Collaboration ISBN: Citation: Towers, A., Sheridan, J., Newcombe, D., & Szabó, Á. (2018). New Zealanders alcohol consumption patterns across the lifespan. Wellington: Health Promotion Agency. Acknowledgements HPA would like to thank those respondents who took the time to participate in this research. Copyright The copyright owner of this publication is HPA. HPA permits the reproduction of material from this publication without prior notification, provided that fair representation is made of the material and HPA is acknowledged as the source. Disclaimer This research has been carried out by an independent party under contract to HPA. The views, observations and analysis expressed in this report are those of the authors and are not to be attributed to HPA. This report has not been externally peer reviewed. This document is available at: Any queries regarding this report should be directed to HPA at the following address: Health Promotion Agency, PO Box 2142, Wellington 6140, New Zealand

3 iv

4 The authors of this report are: Dr Andy Towers (School of Health Sciences, Massey University) Professor Janie Sheridan (School of Pharmacy and the Centre for Addiction Research, the University of Auckland) Dr David Newcombe (School of Population Health and the Centre for Addiction Research, the University of Auckland) Dr Ágnes Szabó (School of Health Sciences, Massey University) v

5 Acknowledgements Researchers involved in the completion of this report New Zealand Health, Work and Retirement Longitudinal Study Research Team We would like to thank Dr Mary Breheny (School of Health Sciences, Massey University) for her leadership of the Life Course History (LCH) project for the Health and Ageing Research Team at Massey University, and for facilitating the inclusion of the alcohol questions upon which this analysis is based. We would like to thank Professors Fiona Alpass and Christine Stephens (School of Psychology, Massey University) who are Principal Investigators of the New Zealand Health, Work and Retirement Longitudinal Study (NZHWR) within which the LCH project was funded and delivered. We would also like to thank Dr Joanne Allen (Research Officer, NZHWR) for her assistance in compiling, scoring and providing access to the NZHWR Life Course History data. International advisers We would like to acknowledge the advice and feedback from our international collaborators: Professor Annie Britton (University College London); Dr Martin Hyde (Swansea University); Professor Alison Moore (University of California, San Diego); Professor Emeritus Christine Savage (Johns Hopkins University); Dr Priscilla Martinez (Alcohol Research Group, University of California, Berkley); Professor Thomas Clausen (University of Oslo); Dr Nadia Minicuci and Dr Ilaria Rocco (National Research Council, Neuroscience Institute, Padova, Italy); and Dr Paul Kowal (World Health Organization). Funding for studies included in the completion of this report The Life Course History Interview study conducted as part of the Health, Work & Retirement Longitudinal Study (New Zealand) Funding for the Life Course History data collection project in 2017 and the wider Health, Work & Retirement Longitudinal Study data collection ( ) was provided by the Ministry of Business, Innovation and Employment (MAUX1403: Enabling Participation By All Older People). HPA would like to thank those respondents who took the time to participate in this research. vi

6 Table of Contents Table of Contents vii Executive summary x 1.0 Literature overview: Alcohol consumption over the lifespan Critical life events predictive of alcohol consumption The life course approach to studying alcohol consumption Aims of the current study Life Course History Interview Study: Method and measures The NZHWR Life Course History Interview Study: Sample and design Measurement of alcohol use The measurement of predictors of alcohol use Analytic plan Results for question 1: Initiation of alcohol use Key findings: Initiation of alcohol use Initiation of alcohol use Results for question 2: Patterns of alcohol use over the lifespan Key findings: Patterns of alcohol use over the lifespan Lifespan drinking profiles of men Lifespan drinking profiles of women Childhood predictors of lifespan drinking profiles Results for question 3: Transitioning from hazardous to non-hazardous drinking Key findings: Transitioning from hazardous to non-hazardous drinking Prevalence of hazardous drinking over the lifespan Lifespan predictors of hazardous drinking Limitations 28 Appendices 29 Appendix 1. The interview materials and measures Appendix 2. The measurement of predictors of alcohol use Appendix 3. Parameter estimates of predictors of alcohol use initiation Appendix 4. Fit statistics for longitudinal drinking trajectory analysis References 36 vii

7 List of Figures Figure 1. Probability of initiating alcohol use between the age of 10 and 23 years Figure 2. Differences in the probability of alcohol use initiation based on gender and Māori descent Figure 3. Differences in the probability of alcohol use initiation based on gender and childhood socioeconomic status Figure 4. Differences in the probability of alcohol use initiation based on gender and parents smoking behaviour Figure 5. Differences in the probability of alcohol use initiation based on self-reported primary school performance Figure 6. Men in Group 1: Low frequency and moderate to low quantity drinking Figure 7. Men in Group 2: Increasing frequency but moderate quantity slowly reducing across life. 15 Figure 8. Men in Group 3: Higher frequency and quantity reducing after mid-life Figure 9. Women in Group 1: Low frequency and low quantity drinking Figure 10. Women in Group 2: High frequency and low quantity drinking Figure 11. Cross-lagged panel model showing reciprocal relationship between employment loss and the likelihood of becoming a hazardous drinker over the lifespan Figure 12. Cross-lagged panel model showing reciprocal relationship between the development of chronic conditions and the likelihood of becoming a non-hazardous drinker over the lifespan Figure 13. Cross-lagged panel model showing reciprocal relationship between relationship status and the likelihood of becoming a hazardous drinker over the lifespan Figure 14. Example page from life history calendar viii

8 List of Tables Table 1. AUDIT-C questions modified for the LCH study Table 2. Descriptive statistics of early life predictors for the total sample and by gender Table 3. Descriptive statistics of mid-life predictors Table 4. Early life characteristics of the lifespan drinking trajectory groups Table 5. Early life predictors of lifespan drinking trajectories of men and women Table 6. Prevalence of hazardous versus non-hazardous drinkers in each decade across the lifespan Table 7. Prevalence of drinkers transitioning between hazardous and non-hazardous drinking patterns in each decade across the lifespan Table 8. Most frequent patterns of hazardous drinking over the lifespan Table 9. Parameter estimates of predictors of alcohol use initiation Table 10. Fit indices for the growth mixture models for men Table 11. Parameter estimates of the intercept, linear and quadratic slopes for frequency and quantity of drinking of men in the 3-profile solution Table 12. Fit indices for the growth mixture models for women Table 13. Parameter estimates of the intercept, linear and quadratic slopes for frequency and quantity of drinking of women in the 2-profile solution ix

9 Alcohol consumption over the lifespan Executive summary Hazardous drinking increases the risk of alcohol-related harm for older drinkers because they are more sensitive to the effects of alcohol, and more likely to have chronic health conditions or use medications that alcohol can interfere with [1]. One way to effectively prevent alcohol-related harm in later life is to better understand the early- and mid-life factors and life events that trigger hazardous drinking. An individual s alcohol use at any point in the lifespan is influenced by a range of environmental and personal factors [2, 3]. Environmental influences include social norms around drinking [4, 5], availability and price of alcohol [6], or opportunities for alcohol consumption [7]. Personal influences include biological markers and genetic vulnerabilities [2, 3], individual differences in personality or temperament [8, 9], or socio-demographic characteristics (e.g., gender, marital status) [10]. Protective (i.e., skills, resources or attributes that eliminate or mitigate risk) and risk factors might be present at different stages of life and influence drinking behaviour. Over the course of the lifespan, people experience different events that might prompt them to increase or decrease their drinking. Aims of this study This study aimed to answer the following questions: Question 1: Initiation of alcohol use What key demographic factors and early childhood life events were associated with early initiation of alcohol use in the current cohort of New Zealanders aged 60 years or older? Question 2: Patterns of alcohol use over the lifespan What are the main patterns of alcohol use of older New Zealanders over the lifespan? What are the key demographic factors and early childhood life events associated with lifespan drinking trajectories of older New Zealanders? Question 3: Transitioning from hazardous to non-hazardous drinking What are the key life events and life transitions that created a context for the development of hazardous alcohol use in older New Zealanders? Methods The data for this study were drawn from the Life Course History (LCH) Interview Study embedded in the New Zealand Health, Work and Retirement Longitudinal Study (NZHWR). A total of 801 NZHWR participants completed interviews. The LCH interviews were 1-2 hour-long computer-assisted telephone interviews, and collected information from participants regarding key early- and mid-life events and circumstances likely to influence their health, wealth and social stability in older adulthood. x

10 Results of the study Question 1: Initiation of alcohol use In an evaluation of the key factors and early childhood life events that were associated with initiation of alcohol use during adolescence and young adulthood in this cohort of older New Zealanders, we found that: 68% had their first drink between the age of 14 and 18, and the majority (88%) before the age of 23 only 4% of the sample were lifetime abstainers (i.e., never consumed alcohol) 82% of men had started drinking by 18 years of age compared to only 56% of women increased likelihood of earlier drinking initiation was related to: o higher childhood socioeconomic status o having a parent that smoked o being of Māori descent reduced likelihood of early initiation was related to having better self-reported educational achievement overall, when all predictor variables were considered together, only gender, socioeconomic status and parents smoking behaviour influenced early initiation. This suggests that any differences in education and ethnicity are likely attributable to underlying differences in socioeconomic status. Question 2: Patterns of alcohol use over the lifespan When we explored the alcohol use patterns (frequency of use and quantity typically consumed on each occasion) of this cohort of older New Zealanders, we found three distinct groups of men and two distinct groups of women who shared common drinking patterns across their lifespan. The three groups of male drinkers were made up of: 1. those who drank alcohol infrequently in moderate-to-low quantities over their lifespan (36%) 2. those who drank alcohol frequently in moderate-to-low quantities over their lifespan (51%) 3. those who drank alcohol frequently and in high quantities over the lifespan (13%). The two groups of female drinkers were made up of: 1. those who drank alcohol infrequently and consumed low quantities at each occasion (48%) 2. those who drank frequently in moderate-to-low quantities over the lifespan (52%). An investigation of the factors associated with these lifespan drinking trajectories found that: men who were frequent drinkers consuming low-to-moderate quantities were more likely to have initiated alcohol use earlier and had better school performance at the age of 10 than infrequent male drinkers women who were high frequency drinkers and consumed low quantities per occasion initiated alcohol use at an earlier age and had more affluent childhood socioeconomic status than infrequent female drinkers xi

11 men who were high frequency drinkers consuming high quantities per occasion initiated alcohol use earlier, had poorer socioeconomic status, and were much more likely to have a heavily drinking parent than infrequent male drinkers. Question 3: Transitioning from hazardous to non-hazardous drinking In terms of the nature of hazardous drinking levels across the lifespan of this cohort of older New Zealanders: drinking patterns were largely stable across lifespan, with long periods of hazardous or nonhazardous drinking being the norm one-third of participants (36%) became hazardous drinkers as adolescents or young adults, and remained hazardous drinkers throughout the lifespan only a small proportion (14%) were life-long (i.e., from adolescence onwards) non-hazardous drinkers transition into or out of hazardous drinking was not common (less than 10% in each decade); when it occurred, it was usually a singular event in the lifespan (i.e., no further transitions occurred). Transitioning in and out of hazardous drinking was linked to three key life events: Unemployment prior to mid-life (i.e., before 40s) was associated with increased likelihood of developing hazardous drinking habits, which, in turn, was associated with further difficulties with finding employment. Relationship breakdown in mid-life was associated with an increased likelihood of hazardous drinking. However, hazardous drinking was not associated with an increased probability of future divorce or separation. Development of a chronic health condition in young adulthood and mid-life was associated with an increased likelihood of transitioning from hazardous to non-hazardous drinking. Conclusion In a sample of over 800 older New Zealanders, initiation of alcohol use during adolescence (from 14 to 18 years of age) was common, more likely in men, and a hallmark of frequent drinking in later life. In men, early initiation was more likely in households with parents that smoked, while in women it was more likely in higher socioeconomic households. For men, better educational performance in childhood was associated with frequent but lower quantity drinking across adulthood, whereas a childhood of economic disadvantage and the presence of heavy-drinking parents was linked to high frequency and high quantity drinking patterns later in life. For women, a childhood of economic advantage predicted high frequency but low quantity drinking. Thirteen percent of the sample were hazardous drinkers across the lifespan. Change from hazardous to non-hazardous drinking (or vice versa) was uncommon (i.e., less than 10% in each decade). If it did occur, further transitions were unlikely. The stability of drinking in this cohort suggests that, once established, a pattern of drinking (whether hazardous or non-hazardous) is unlikely to be modified and becomes a stable trait. This indicates that the likelihood of normalising hazardous drinking patterns is at the point of initiation in adolescence and early adulthood, and thus the critical point of intervention to reduce alcohol-related harm. xii

12 Only a small minority of participants showed movement between hazardous and non-hazardous drinking patterns across their lives, which appeared to be indicative of key early- to mid-life events. Unemployment between ages 20 and 40 and loss of a relationship between ages 30 and 50 increased the risk of changing from a non-hazardous to a hazardous drinker. Conversely, developing a chronic health condition before 50 years of age increased the likelihood of transitioning from hazardous to non-hazardous consumption. These results indicate that middle adulthood (from 30 to 50) is the period when change in drinking behaviour may occur and for very specific reasons. This is the time when people experience the most financial challenges (e.g., raising children, paying off mortgage) and greatest responsibilities, both in their personal (looking after children and ageing parents) and professional lives (career development). Separation and unemployment present major financial strains, which have been found to predict hazardous drinking [11]. At the same time, when people acquire chronic diseases earlier in life, they might be more motivated to make positive lifestyle changes. Middle adulthood (from 30 to 50) is an age-group that is rarely considered to be at risk, and therefore, rarely targeted by interventions or policies. Overall, drinking behaviour in old age mirrors drinking habits developed in young adulthood. While alcohol consumption is generally stable over the lifespan, difficult life events experienced mid-life are likely to prompt significant changes in alcohol use. This highlights the importance for policies to promote low-risk alcohol drinking and reduction of use across the lifespan and for health care providers to pay particular attention to individuals experiencing difficult life transitions so that targeted interventions can take place if needed. xiii

13 1.0 Literature overview: Alcohol consumption over the lifespan The World Health Organization (WHO) states that hazardous drinking is a pattern of alcohol consumption that increases the risk of harmful consequences for the user or others. Hazardous drinking patterns are of public health significance despite the absence of any current disorder in the individual user (p. 1) [12]. Research shows that over one-third of New Zealanders aged 55 years or older are hazardous drinkers, and that they drink more frequently and in higher quantities than older adults in many other countries [13]. Hazardous drinking increases the risk of alcohol-related harm for older drinkers because they are more sensitive to the effects of alcohol, and more likely to have chronic health conditions or use medications that alcohol can interfere with [1]. One way to effectively prevent alcohol-related harm in later life is to better understand the early- and mid-life factors and life events that trigger hazardous drinking. An individual s alcohol use at any point in the lifespan is influenced by a range of environmental and personal factors [2, 3]. Environmental influences include social norms around drinking [4, 5], availability and price of alcohol [6], or opportunities for alcohol consumption [7]. Personal influences include biological markers and genetic vulnerabilities [2, 3], individual differences in personality or temperament [8, 9], or socio-demographic characteristics (e.g., gender, marital status) [10]. Protective (i.e., skills, resources and attributes that mitigate or eliminate risk) and risk factors might be present at different stages of life and influence drinking behaviour. Over the course of the lifespan, people experience different events that might prompt them to increase or decrease their drinking Critical life events predictive of alcohol consumption Childhood and adolescence Knowledge of a person s early experiences with alcohol is crucial to understanding drinking behaviour later in life. Adverse childhood experiences, such as parental alcohol abuse, parental divorce or physical abuse, have been linked, not only to early onset of alcohol use [14-16], but also to hazardous drinking later in life [17]. For example, using retrospective life history interviews in the British Whitehall II study, Leung, Britton, and Bell [18] found that exposure to multiple adverse life experiences (e.g., parental divorce, parental mental health/alcohol problems, physical abuse) was associated with a greater likelihood of hazardous drinking in mid- and late-life. Further research highlights wider risk factors for the early onset of alcohol use, such as poor educational achievement [19-21] and parental health behaviours [22]. The effect of childhood socioeconomic status is inconclusive. Systematic reviews have found little robust evidence for any association between childhood socioeconomic status and alcohol use during adolescence [23, 24] or adulthood [25]. Numerous longitudinal and retrospective studies highlight that early onset of drinking significantly increases the odds of alcohol abuse and dependence in adulthood [26-30]. However, Guttmannova et al. [31] suggested that the impact of early onset of alcohol use on later life alcohol consumption patterns might reflect other individual, family and social risk factors that make people vulnerable for adult alcohol problems. Their analyses using 25-year longitudinal data from the United States Seattle Social Development Project showed that early onset of drinking was associated with dependence in young adulthood, but there was no relationship with lifelong alcohol use patterns after controlling 1

14 for socio-demographic characteristics and other substance use. In this respect, while there is compelling evidence for a link between early age of drinking onset and later life drinking patterns, such a relationship is likely founded on the existence of wider social and environmental risk factors Adulthood Stressful life events are important factors influencing drinking behaviour over the course of adulthood [32, 33]. Job loss, for example, is considered a major catalyst of negative health behaviours. Research in multiple countries and in workers of various ages indicates that hazardous drinking is much more likely to occur after losing employment [34-36], in some cases increasing that likelihood by up to four times [37]. A systematic review of published literature on unemployment and substance use also indicates that employment reduces the likelihood of being a hazardous drinker [38]. In general, this research suggests a bidirectional relationship, such that hazardous drinking increases the likelihood of job loss, whereas unemployment acts as a risk factor for hazardous drinking. Another stressful life event commonly investigated in relation to hazardous drinking is divorce or separation. Numerous studies link divorce to an increased risk of heavy or hazardous drinking [39-42], although the effect might vary by gender [43]. Research findings regarding the direction of the effect are less conclusive, with some studies suggesting that heavy drinking is not only a consequence of but also a contributing factor to divorce [44]. While some life events increase the risk of hazardous drinking, others may prompt individuals to reduce their drinking. For example, the onset of illness of poor health is often reported as a major reason to stop or reduce alcohol consumption [45-47]. Across Europe and the United States, the development of medical conditions (e.g., cancer) and acute health events (e.g., joint aches) [48] or poor self-rated health [34] predicted abstinence or a reduction in alcohol use in older adulthood. In sum, research suggests that the relationship between life transitions and alcohol consumption is complex, with some events (e.g., unemployment) enhancing risk of hazardous drinking, while others (e.g., acquisition of illnesses) lead to reduced consumption The life course approach to studying alcohol consumption The life course perspective provides a useful framework for investigating key life events and stage transitions that influence alcohol consumption in older New Zealanders. Life course history studies can capture period and cohort effects. For example, Grant [49] demonstrated that in young adults born in the United States before World War II, fewer than 50% reported that they consumed alcohol. In contrast, 75% of young adults born after the Vietnam era ( ) were drinkers. Further, life course studies allow researchers to identify sensitive periods of alcohol use and the potentially harmful impact of drinking on health over time. New Zealand currently lacks the large-scale, birth cohort studies with samples of sufficient age with which to assess the effects of early- and mid-life events on alcohol use trends in later life. When such longitudinal data are not available from early childhood, the only way to gather information about lifespan experiences is through retrospective data collection using life course history interviews. Life course history interviews elicit and record information about a range of factors, which may affect health and wellbeing during the lifespan. Structured life course history interviews have been developed for other studies of ageing, including the English Longitudinal Study on Ageing (ELSA), which administered interviews face-to-face, and the Australian 45 and Up Study, which administered interviews over the telephone. 2

15 2.0 Aims of the current study The objective of the present study was to investigate how exposure to critical life events affects alcohol use over the lifespan and later life health outcomes in older New Zealanders. Using the Life Course History (LCH) Interview Study embedded in the New Zealand Health, Work and Retirement Longitudinal Study (NZHWR), this study aimed to answer the following questions: Question 1: Initiation of alcohol use What key demographic factors and early childhood life events were associated with early initiation of alcohol use in the current cohort of New Zealanders aged 60 years or older? Question 2: Patterns of alcohol use over the lifespan What are the main patterns of alcohol use of older New Zealanders over the lifespan? What are the key demographic factors and early childhood life events associated with lifespan drinking trajectories of older New Zealanders? Question 3: Transitioning from hazardous to non-hazardous drinking What are the key life events and life transitions that created a context for the development of hazardous alcohol use in older New Zealanders? 3

16 3.0 Life Course History Interview Study: Method and measures 3.1 The NZHWR Life Course History Interview Study: Sample and design The NZHWR s LCH Study is an initiative of Massey University s Health & Ageing Research Team (HART). The LCH interviews were based on the content and protocols utilised in the Survey of Health, Ageing and Retirement in Europe [50], which was conducted to better understand the lifetime determinants of health and wellbeing in later life in 19 countries across Europe. Interview participants for the LCH Project were drawn from existing participants in the NZHWR, a national survey of healthy ageing in New Zealanders aged 55 and over. NZHWR participants were first recruited in 2006 and have been surveyed by post on a biennial basis [51]. Appendix 1 provides a comprehensive overview of the design of the LCH interviews used in this study. Briefly, these were 1-2 hour-long computer-assisted telephone interviews using a standard protocol to collect information from participants regarding key early- and mid-life events and circumstances likely to influence their health, wealth and social stability in older adulthood. A total of 1,133 NZHWR participants who were still active in the study in 2014 were approached to be interviewed in the LCH study. Of these, 236 declined and 96 were unable to be contacted during the follow-up, resulting in 801 completed interviews. Participant ages ranged from 61 to 81 (M = 72, SD = 4.5). Approximately 40% of respondents were of Māori descent, and a bit more than half of the sample were female (52%). 3.2 Measurement of alcohol use The LCH interview schedule included a retrospective alcohol life-grid [34] originally employed in the Whitehall II study of British civil servants to ascertain the patterns of drinking frequency, quantity and bingeing in each decade of the lifespan. Professor Annie Britton of University College London (Primary Investigator on the Whitehall II study and author of the life-grid) was an advisor on the use of this life-grid in the NZHWR LCH and in the analysis of the alcohol data. The life-grid comprised two specific components. First, a question identifying whether the participant had ever consumed alcohol, and if so a further question identifying the age of first consumption. Second, participants pattern of drinking frequency, quantity and binge drinking were assessed with the AUDIT-C items modified for the LCH study to assess alcohol use across decades of life (Table 1). Research from the British Whitehall II study shows that this retrospective alcohol lifegrid offers a reliable assessment of alcohol use across the lifespan for older adults [34]. Specifically, participants were requested to identify the typical frequency of alcohol use, the typical quantity of use per occasion, and the typical regularity of consuming 6-or-more alcoholic drinks on any single occasion for the following time periods up until their current decade of life: years, years, years, years, years, years, years, 80+ years. As per standard scoring for the AUDIT-C [52], a composite score was created by summing responses on the three indicators to provide a total ranging from A score of 3 or higher for women and a score of 4 or higher for men indicated hazardous levels of drinking. 4

17 Table 1. AUDIT-C questions modified for the LCH study. 1. Frequency: How often did you have a drink containing alcohol when you were years old? (0 = Never, 1 = Monthly or less, 2 = 2-4 times per month, 3= 2-3 times per week, 4 = 4+ times per week) 2. Quantity: How many drinks containing alcohol did you have on a typical occasion when you were years old? (0 = 1-2 drinks, 1 = 3-4 drinks, 2 = 5-6 drinks, 3 = 7-9 drinks, 4 = 10+ drinks) 3. Binge drinking: How often did you have six or more drinks on one occasion when you were years old? (0 = Never, 1 = Less than monthly, 2 = Monthly, 3 = Weekly, 4 = Daily or almost daily) 3.3 The measurement of predictors of alcohol use What follows is a brief indication of the range of measures of childhood and mid-life predictors of alcohol use utilised in this study. Appendix 2 offers a more comprehensive overview of these measures. Descriptive statistics for the early- and mid-life predictors are reported in Table 2 and Table 3 respectively Potential early-life predictors of alcohol initiation and use Childhood material wellbeing: The LCH study used items adapted from the New Zealand Deprivation Index. These items assessed specific aspects of material wellbeing in the participant s life at the age of 10, including living conditions and household facilities, family ownership of household items, the type of school attended, opportunities for social participation, and access to food. A childhood socioeconomic status index was created with a range of 0 to 23 (Median = 19). Higher scores indicate better economic wellbeing. The total score was further divided to indicate low (0-12), medium (13-18), and high (19-23) childhood socioeconomic status (SES). Childhood health history: Participants were asked to indicate whether they had any of a range of illnesses or health conditions during childhood, including polio, asthma, allergies (other than asthma), or infectious diseases. A composite health score was created by summing all items (range: 0-9; Median = 2). Higher scores indicate more chronic health conditions in childhood. Adverse childhood experiences: Based on research by Leung et al. [18], participants were asked about stressful life events experienced during childhood: o physical abuse, o sexual abuse, o hospitalisation for four weeks or longer, o parental divorce, o parental alcohol abuse, o and having a parent with mental health problems. A composite score was created by summing all items (range: 0-5; Median = 1). Higher scores indicate more adverse childhood experiences. Particular events, such as parental alcohol abuse, were investigated separately in some analyses. Parents smoking behaviour: Participants indicated whether during their childhood, any of their parents or guardians smoked (Yes/No). 5

18 Childhood educational performance: Participants were asked to think back to their time in school when they were 10 years old and rate their performance in English compared to other children in their class. Responses were divided into two categories: 1) same as or better than peers, and 2) worse than peers Potential mid-life predictors of alcohol use Employment history: Employment history was mapped onto each decade of life starting from the teenage years up until the 70s by differentiating between continuous employment (a period of 6-months or longer) and unemployment. This allowed us to identify whether and in which decade of life a person became unemployment for at least six months. Adult health history: Participants were asked to describe periods of ill health or disability that lasted for more than a year, or serious illnesses that lasted less than one year but influenced daily life for more than a year (e.g., cancer). Relationship history: Participants indicated how many times they have been married or lived unmarried with someone else as a couple. For each past relationship, participants were asked to report the year in which they separated from their partner. 3.4 Analytic plan Research question 1: Initiation of alcohol use Survival analysis was used to estimate the probability of initiating alcohol use between the age of 10 and 23. Time to initiation was measured in years and multivariate, discrete-time hazard regression analysis was used to model predictors of the progression to alcohol initiation. Predictor variables tested were gender, ethnicity, childhood socioeconomic status, childhood health conditions, childhood adverse experiences, parental health behaviours (i.e., smoking and alcohol use), and childhood educational performance Research question 2: Patterns of alcohol use over the lifespan We used Growth Mixture Modelling to test the second research question. This data analytic technique aims to identify subgroups of individuals who demonstrate similar change trajectories over time [53]. We modelled changes in quantity and frequency of drinking over the lifespan based on consumption levels reported for each decade from adolescence (16-19 years of age) through to old adulthood (70s). We conducted the analysis for men and women separately to explore alcohol use trajectories specific to each gender Research question 3: Transition from hazardous to non-hazardous drinking Cross-lagged analysis was performed to examine whether changes in employment, adult health and relationship status were predictive of transitioning in and out of hazardous drinking over the lifespan. Transitions were measured decade by decade, starting with the adolescent years up until the late 60s. This model estimated bi-directional, causal influences between hazardous drinking and the three life events (i.e., unemployment, relationship breakdown, and acquisition of chronic health conditions) over time. The model also provided information about the stability of these relationships over the lifespan. 6

19 Table 2. Descriptive statistics of early life predictors for the total sample and by gender. Gender Total sample Female Male N % N % N % Female Male Māori descent Not of Māori descent Māori descent Childhood SES Low to mid High No. of chronic illnesses None One Two or more No. of adverse childhood experiences None One Two or more Physical abuse before the age of 10 No Yes Sexual abuse before the age of 10 No Yes Hospitalisation for >4 weeks No Yes Parental alcohol abuse No Yes Parental mental illness No Yes Parental divorce Parents together Divorced Parental smoking No Yes Educational performance Same as or better than peers Worse than peers

20 Table 3. Descriptive statistics of mid-life predictors. From 10s to 20s From 20s to 30s From 30s to 40s From 40s to 50s From 50s to 60s From 60s to 70s N % N % N % N % N % N % Employment history Continuous % % % % % % Lost employment % % % % % % Health history No new illness % % % % % % Acquired an illness % % % % % % Relationship history Remained in relationship % % % % % % Separated 6 0.7% % % % % % 8

21 4.0 Results for question 1: Initiation of alcohol use 4.1 Key findings: Initiation of alcohol use In an evaluation of the key factors and early childhood life events that were significantly associated with initiation of alcohol use during adolescence and young adulthood, we found that: 68% had their first drink between the age of 14 and 18, and the majority (88%) before the age of 23 only 4% of the sample were lifetime abstainers (i.e., never consumed alcohol) 82% of men had started drinking by 18 years of age compared to only 56% of women increased likelihood of earlier drinking initiation was related to: o higher childhood socioeconomic status o having a parent that smoked o being of Māori descent reduced likelihood of early initiation was related to having better self-reported educational achievement overall, when all predictor variables were considered together, only gender, socioeconomic status and parents smoking behaviour influenced early initiation. This suggests that any differences in education and ethnicity are likely attributable to underlying differences in socioeconomic status. 4.2 Initiation of alcohol use First, we modelled the probability of initiating alcohol use between the age of 10 and 23 in the overall sample and by gender. Next, we examined demographic and early childhood covariates of alcohol use initiations (parameter estimates are reported in Table 9 in Appendix 3). Figure 1 overleaf illustrates the probabilities of initiating alcohol use for the total sample and by gender. In the total sample, the probability of initiating alcohol use was relatively low until the age of 14 (increasing from 0-7%), at which point there was a steep increase in alcohol use initiation. By the age of 18, 68% of the sample had consumed their first drink. The probability of initiating alcohol use continued to increase after the age of 18, but at a much slower rate. At the age of 23, 88% of participants reported alcohol use. Analyses indicated significant gender differences in the initiation of alcohol use. By the age of 14, almost 1-in-5 men had used alcohol compared with less than 1-in-20 women. By 18 years of age, more than 80% of men but only 56% of women had initiated alcohol use. At the age of 23, 5% of men and 18% of women were still abstainers. 9

22 Probability of initiating alcohol use Probability of initiating alcohol use Figure 1. Probability of initiating alcohol use between the age of 10 and 23 years Total Male Female before Participant age (years) Individual predictors of alcohol use initiation Overall no significant differences were found between participants of Māori and non-māori descent (Figure 2). Figure 2. Differences in the probability of alcohol use initiation based on gender and Māori descent Non-Maori Women Maori Women Non-Maori Men Maori Men before Participant age (years) 10

23 Probability of initiating alcohol use The results in Figure 3 illustrate that childhood socioeconomic status was a significant predictor of alcohol initiation. Specifically, those growing up with better economic conditions were more likely to initiate alcohol use at any age, but especially after the age of 18. Figure 3. Differences in the probability of alcohol use initiation based on gender and childhood socioeconomic status Women - High SES Women - Low to mid SES Men - High SES Men - Low to mid SES 0 before Participant age (years) 11

24 Probability of initiating alcohol use Adverse childhood experiences as predictors of alcohol use initiation were tested individually, but none of the life events had significant effects. Nor did the composite score. This suggests that the accumulation of stressful childhood life events did not significantly influence the onset of alcohol use in this sample. However, significant effects were found based on parents smoking behaviours. Those with a parent who smoked were more likely to start using alcohol after the age of 16 compared with those whose parents were non-smokers (Figure 4). Figure 4. Differences in the probability of alcohol use initiation based on gender and parents smoking behaviour Women - Non-smoking parents Women - Smoking parents Men - Non-smoking parents Men - Smoking parents before Participant age (years) 12

25 Probability of initiating alcohol use Finally, educational performance was examined as a predictor of alcohol use initiation. Results (Figure 5) indicated that better self-reported educational performance functioned as a protective factor. Those who reported worse performance in English than their peers at the age of 10 were more likely to initiate alcohol use at any age. Figure 5. Differences in the probability of alcohol use initiation based on self-reported primary school performance Female - worse than peers Female - better than peers Male - worse than peers Male - better than peers 0 before Participant age (years) Simultaneous predictors of alcohol use initiation In a separate analysis, we simultaneously evaluated the effect of early childhood predictor variables on the age of alcohol use initiation. The results indicated that gender, socioeconomic status, and parents smoking behaviour were all significantly associated with alcohol use initiation. However, self-reported educational performance was no longer a significant predictor when included alongside other predictors. 1 This suggests that differences in educational performance are likely to reflect differences in socioeconomic status. 1 Critical ratios for the simultaneous predictors of alcohol use initiation: gender (C.R. = 8.80, p <.001); socioeconomic status (C.R. = 3.97, p <.001); parents smoking behaviour (C.R. = 2.03, p =.043), educational performance (C.R. = -1.30, ns). 13

26 5.0 Results for question 2: Patterns of alcohol use over the lifespan 5.1 Key findings: Patterns of alcohol use over the lifespan When we explored the alcohol use patterns of older New Zealanders (frequency of use and quantity typically consumed on each occasion), we found three distinct groups of men and two distinct groups of women who shared common drinking patterns across their lifespan. The three groups of male drinkers were made up of: 1. those who drank alcohol infrequently in moderate-to-low quantities over their lifespan (36%). 2. those who drank alcohol frequently in moderate-to-low quantities over their lifespan (51%). 3. those who drank alcohol frequently and in high quantities over the lifespan (13%). The two groups of female drinkers were made up of: 1. those who drank alcohol infrequently and consumed low quantities at each occasion (48%) 2. those who drank frequently in moderate-to-low quantities over the lifespan (52%). An investigation of the predictors of these lifespan drinking trajectories found that: men who were frequent drinkers consuming low-to-moderate quantities were more likely to have initiated alcohol use earlier and had better school performance at the age of 10 than infrequent male drinkers. women who were high frequency drinkers and consumed low quantities per occasion initiated alcohol use at an earlier age and had better childhood socioeconomic status than infrequent female drinkers. men who were high frequency drinkers and consuming high quantities per occasion initiated alcohol use earlier, had poorer socioeconomic status, and were much more likely to have a heavily drinking parent than infrequent male drinkers. those with a heavily drinking parent were almost five times more likely to become frequent heavy drinkers than infrequent, low quantity consumers, and almost three times more likely to become frequent heavy drinkers than frequent consumers of moderate quantity of alcohol. This finding provides further support for the influence of parents health behaviours on their children s alcohol use practices [22]. 5.2 Lifespan drinking profiles of men Based on frequency and quantity of alcohol consumption throughout the decades of life, we explored whether there were distinct profiles of men and women who shared the same (or similar) patterns of drinking across the lifespan. Our statistical models (see Tables in Appendix 4) identified three distinct trajectories of drinking across the lifespan in men and two distinct trajectories in women. The first group of men (n = 131; 35.8%) was characterised by low frequency drinking throughout the lifespan (i.e., on average consuming alcohol monthly or less or 2-4 times per month ). Quantity of alcohol consumption in this group was the highest in the teenage years (around 3-4 drinks per occasion), and slightly declined over time (Figure 6 over). 14

27 Figure 6. Men in Group 1: Low frequency and moderate to low quantity drinking. In the second group of men (n = 188; 51.4%), frequency of alcohol consumption increased from the teenage years (i.e., 2-4 times a month) until the 50s (2-3 times per week), plateaued between the 50s and 60s, and slightly declined after the 60s. The quantity consumed was at its highest in the teenage years (i.e., 3-4 drinks per occasion), and slightly declined over time (Figure 7). Figure 7. Men in Group 2: Increasing frequency but moderate quantity slowly reducing across life. The third and smallest group of men (n = 47; 12.8%) displayed patterns of frequent heavy drinking. Frequency of alcohol use increased from the teenage years until the 30s (i.e., from 2-4 times a month to 2-3 times per week), plateaued between the 30s and 40s, and showed a sharp decline after the 40s. Quantity of alcohol consumed at each occasion was already high in the teenage years (i.e., 5-6 drinks per occasion), which further increased until the 30s and remained stable between the 30s and 40s (i.e., between 5-6 and 7-9 drinks per occasion). Although a sharp decline was observed after the 40s, member of this group still reported drinking 3-4 standard drinks per occasion in their 70s (Figure 8 over). 15

28 Figure 8. Men in Group 3: Higher frequency and quantity reducing after mid-life. 5.3 Lifespan drinking profiles of women The first group of women (n = 183; 47.8%) reported low frequency drinking (i.e., monthly or less) and low quantity per occasion (i.e., 1-2 drinks) drinking throughout the lifespan, with no significant change over time (Figure 9). Figure 9. Women in Group 1: Low frequency and low quantity drinking. 16

29 In the second and slightly larger group (n = 200; 52.2%), the frequency of alcohol consumption increased throughout the lifespan (from monthly or less to 2-3 times a week ) until the 60s, and remained stable afterwards. The quantity of alcohol consumed on each occasion was moderate to low and slightly decreasing (i.e., between 1-2 drinks per occasion and 3-4 drinks per occasion in the teenage years and early 20s, after which it slowly declined to 1-2 drinks per occasion; Figure 10). Figure 10. Women in Group 2: High frequency and low quantity drinking. 5.4 Childhood predictors of lifespan drinking profiles Next, we examined the association between childhood experiences and lifespan drinking patterns of men and women. A multinomial logistic regression analysis was conducted to test predictors of being assigned to one of the three drinking trajectory groups in men. For women, we employed a binomial logistic regression analysis because there were only two trajectory groups. Predictor variables included in the models were: (1) the age of onset of drinking, (2) childhood socioeconomic status, (3) the number of childhood chronic illnesses, (4) educational performance at the age of 10, (5) hospitalisation for four weeks or longer, (6) a parent who drank heavily, (7) having a parent with mental health problems, and (8) parental divorce. Physical and sexual abuse could not be included as predictor variables because of the low incidence rates. The demographic and early-life characteristics of each of the three groups for men and the two groups for women are displayed in Table 4 over. The extent to which these characteristics were associated with drinking trajectory groups is displayed in Table 5 over. Childhood predictors of lifespan drinking profiles of men Comparing Group 2 (Increasing frequency but moderate-to-low quantity) to Group 1 (Low frequency and quantity): Men in the Group 2 initiated alcohol use earlier, had better selfreported school performance at the age of 10, and were less likely to have a parent with mental health problems than men in Group 1. Comparing Group 3 (High frequency and quantity) to Group 1 (Low Frequency and Quantity): Men in Group 3 initiated alcohol use earlier, had poorer socioeconomic status, and were much more likely to have a heavily drinking parent than men in Group 1. 17